### PPT

```KEY DERIVATION
WITHOUT
ENTROPY WASTE
Yevgeniy Dodis
New York University
Based on joint works with B. Barak, H. Krawczyk, O. Pereira, K. Pietrzak, F-X. Standaert, D. Wichs and Y. Yu
Key Derivation
2

Setting: application P needs m–bit secret key R

Theory: pick uniformly random R  {0,1}m

Practice: have ”imperfect randomness” X  {0,1}n
 physical
sources, biometric data, partial key leakage,
extracting from group elements (DH key exchange), …

Need a “bridge”: key derivation function (KDF)
h: {0,1}n  {0,1}m s.t. R = h(X) is “good” for P
…
only assuming X has “minimal entropy” k
Formalizing the Problem
3

Ideal Model: pick uniform R  Um as the key
 Assume

P is e–secure (against certain class of attackers A)
Real Model: use R = h(X) as the key, where
= H(X) ≥ k (Pr[X = x] 2− , for all x)
Real
Security
e’

Ideal
Security
e
n
m
 h: {0,1}  {0,1} is a (possibly probabilistic) KDF
 min-entropy(X)

Goal: minimize k s.t. P is 2e–secure using R = h(X)
minimize entropy loss L = k  m
 (If possible, get information-theoretic security)
 Note: we design h but must work for any (n, k)-source X
 Equivalently,
Formalizing the Problem
4

Ideal Model: pick uniform
R  Um as the key
h
 Assume

X
P is e–secure (against certain class of attackers A)
Real Model: use R = h(X) as the key, where
= H(X) ≥ k (Pr[X = x] 2− ,Xfor all x)
 h: {0,1}n  {0,1}m is a (possibly probabilistic) KDF
 min-entropy(X)

Goal: minimize k s.t. P is 2e–secure using R = h(X)
minimize entropy loss L = k  m
 (If possible, get information-theoretic security)
 Note: we design h but must work for any (n, k)-source X
 Equivalently,
Old Approach: Extractors
5

Tool: Randomness Extractor [NZ96].
 Input:
a weak secret X and a uniformly random seed S.
 Output: extracted key R = Ext(X; S).
 R is uniformly random, even conditioned on the seed S.
(Ext(X; S), S) ≈ (Uniform, S)

Many uses in complexity theory and cryptography.
 Well
beyond key derivation (de-randomization, etc.)
secret: X
seed:
S
extracted key:
Ext
R
Old Approach: Extractors
6

Tool: Randomness Extractor [NZ96].
 Input:
a weak secret X and a uniformly random seed S.
 Output: extracted key R = Ext(X; S).
 R is uniformly random, even conditioned on the seed S.
(Ext(X; S), S) ≈ (Uniform, S)

(k,e)-extractor: given any secret (n,k)-source X,
outputs m secret bits “e–fooling” any distinguisher D:
statistical distance
| Pr[D(Ext(X; S), S) =1] – Pr[D(Um, S) =1] |  e
Extractors as KDFs
7


Lemma: for any e-secure P needing an m–bit key,
(k,e)-extractor is a KDF yielding security e’ ≤ 2e
LHL [HILL]: universal hash functions are (k,e)-extractors
where k = m + 2log(1/e)

Corollary: For any P, can get entropy loss 2log(1/e)

RT-bound [RT]: for any extractor, k  m + 2log(1/e)
 entropy
…
loss 2log(1/e) seems necessary 
or is it?
Side-Stepping RT
8

Do we need to derive statististically random R?


No for many (most?) other applications P  !
Series of works “beating” RT [BDK+11,DRV12,DY13,DPW13]
Punch line: Difference between
Extraction and Key Derivation !
New Approach/Plan of Attack
9

Step1. Identify general class of applications P
which work “well” with any high-entropy key R


Interesting in its own right !
Step2. Build good condenser: relaxation of
extractor producing high-entropy (but nonuniform!) derived key R = h(X)
Unpredictability Applications
10

Sig, Mac, OWF, … (not Enc, PRF, PRG, …)

Example: unforgeability for Signatures/Macs
Entropy
 Assume: Pr[A forges with uniform key] ≤ e (= negl)
deficiency
 Hope: Pr[A forges with high-entropy key] ≤ e’

Lemma: for any e-secure unpredictability appl. P,
H(R) ≥  −   e’ ≤ 2 e

E.g., random R except first bit 0  e’ ≤ 2e
Plan of Attack
11
 Step1. Argue any unpredictability applic. P

works well with (only) a high-entropy key R


H(R) ≥  −   e’ ≤ 2 e
Step2. Build good condenser: relaxation of
extractor producing high-entropy
(but non-uniform!) derived key R = h(X)
Randomness Condensers
random
12

(k,d,e)-condenser: given (n, k)-source X, outputs m
bits R “e–close” to some (m, m−d)-source Y :
(Cond(X; S), S) ≈e (Y, S) and H(Y | S) ≥ m – d

Cond + Step1  e’ ≤ (1 + 2 )  e

Extractors: d = 0 but only for k  m + 2log(1/e)

Theorem [DPW13]: d = 1 with k = m + loglog(1/e) + 4
 KDF:
log(1/e)-independent hash function works!

Unpredictability Extractors
13

Theorem: provably secure KDF with entropy loss
loglog(1/e) + 4 for all unpredictability applications
 call
such KDFs Unpredictability Extractors
Example: CBC-MAC, e = 2-64, m = 128
LHL: k = 256
;
Now: k = 138
Indistinguishability Apps?
14
 Still, similar plan of attack:

 Step1.
Identify sub-class of indist. applications P
which work well with (only) a high-entropy key R
 Weaker, but still useful, inequality: e’ ≤ e(2 − 1)

Bonus: works even with “nicer” Renyi entropy
Build good condensers for Renyi entropy
 Much simpler: universal hashing still works !
 Step2.
Square-Friendly Applications
Hermitage State
Museum
15
See [BDK+11,DY13] for (natural) definition…
 All unpredictability applications are SQF
 Non-SQF applications: OTP, PRF, PRP, PRG 
 [BDK+11,DY13]: many natural indistinguishability
applications are square-friendly !



CPA/CCA-enc, weak PRFs, q-wise indep. hash, …
End Result (LHL’): universal hashing is provably
secure SQF-KDF with entropy loss log(1/e)
Square-Friendly Applications
Hermitage State
Museum
16
See [BDK+11,DY13] for (natural) definition…
 All unpredictability applications are SQF
 Non-SQF applications: OTP, PRF, PRP, PRG 
 [BDK+11,DY13]: many natural indistinguishability
 Example: CBC Encryption, e = 2-64, m = 128
applications are square-friendly !



LHL:
k = 256
; LHL’:
k = 192hash, …
CPA/CCA-enc,
weak PRFs,
q-wise indep.
End Result (LHL’): universal hashing is provably
secure SQF-KDF with entropy loss log(1/e)
Efficient Samplability?
17

Theorem [DPW13]: efficient samplability of X
does not help to improve entropy loss below

2log(1/e) for all applications P (RT-bound)
 Affirmatively
resolves “SRT-conjecture” from [DGKM12]

log(1/e) for all square-friendly applications P

loglog(1/e) for all unpredictability applications P
Computational Assumptions?
18
Theorem [DGKM12]:,DPW13]: SRT-conjecture 
efficient Ext beating RT-bound for all
computationally bounded D  OWFs exist
 How far can we go with OWFs/PRGs/…?



One of the main open problems
Current Best [DY13]: “computational” extractor
with entropy loss 2log(1/e)  log(1/eprg)

“Computational” condenser?
Summary

Difference between extraction and KDF

loglog(1/e) loss for all unpredictability apps

log(1/e) loss for all square-friendly apps
(+ motivation to study “square security”)

Efficient samplability does not help


Good computational extractors require OWFs


Main challenge: better “computational” KDFs
```